Radioembolization-Induced Changes in Hepatic [18F]FDG Metabolism in Non-Tumorous Liver Parenchyma

Background: [18F]FDG-PET/CT is increasingly used for response assessments after oncologic treatment. The known response criteria for [18F]FDG-PET/CT use healthy liver parenchyma as the reference standard. However, the [18F]FDG liver metabolism results may change as a result of the given therapy. The aim of this study was to assess changes in [18F]FDG liver metabolism after hepatic 90Y resin radioembolization. Methods: [18F]FDG-PET/CT scans prior to radioembolization and one and three months after radioembolization (consistent with the PERCIST comparability criteria), as well as 90Y-PET/CT scans, were analyzed using 3 cm VOIs. The FDG activity concentration and absorbed dose were measured. A linear mixed-effects logistic regression model and logistic mixed-effects model were used to assess the correlation between the FDG-activity concentration, absorbed dose, and biochemical changes. Results: The median SULVOI,liver at baseline was 1.8 (range = 1.2–2.8). The mean change in SULVOI,liver per month with an increase in time was 0.05 (95%CI 0.02–0.09) at p < 0.001. The median absorbed dose per VOI was 31.3 Gy (range = 0.1–82.3 Gy). The mean percent change in ΔSULVOI,liver for every Gy increase in the absorbed dose was –0.04 (95%CI −0.22–0.14) at p = 0.67. The SULblood and SULspleen results showed no increase. Conclusions: The [18F]FDG metabolism in the normal liver parenchyma is significantly but mildly increased after radioembolization, which can interfere with its use as a threshold for therapy response.


Introduction
More and more [ 18 F]FDG-PET/CT studies are being performed for response evaluations, as anatomical imaging modalities lack adequate response assessment ability or useful and consistent response criteria, or have minor prognostic value [1]. In light of these limitations of anatomical imaging, two distinct response assessment criteria were defined for [ 18 F]FDG-PET: the EORTC criteria, in use since 1999 [2]; and the PERCIST criteria, in use since 2009 [1]. Their differences are outlined in Supplementary Table S1. The mean and maximum standard uptake values (SUV mean and SUV max ) corrected for body weight in a 2D region of interest (ROI) in the most metabolic part of as many tumors as possible are used in EORTC [2], while SUVs corrected for lean body mass (=SUL) in a stringent volume of interest (VOI) of 1 cm 3 centered on the most metabolic active part of up to 5 tumors (=SUL peak ) are used in PERCIST [1]. Additionally, PERCIST uses a SUL mean assessment of the right lobe of the liver (or blood pool) for [ 18 F]FDG-PET comparability at different time points and as a threshold to define the response.
In light of the PERCIST criteria and its use of a VOI in the right liver lobe to assess the comparability of [ 18 F]FDG-PET, multiple test-retest studies in healthy volunteers or former cancer patients showed that the quantification of the physiological activity concentrations of [ 18 F]FDG were similar over longer periods of time, including in the liver parenchyma [3]. However, studies performed in oncologic patients (mainly lymphoma patients) showed that some types of chemotherapy had an effect on the healthy liver parenchyma, leading to an increase in mean liver activity upon [ 18 F]FDG-PET [4,5]. This poses a problem with the current response assessment criteria used in lymphoma patients, the Deauville criteria [6][7][8][9][10], in which the response is based on a visual assessment, comparing the tumor uptake to the liver uptake. Theoretically, the response after treatment could be wrongfully overestimated.
For the treatment of hepatic malignancies, radioembolization has gained ground in the last decades. During that time, multiple studies have used [ 18 F]FDG-PET/CT for response assessments and applied the PERCIST criteria [11,12]. However, the concomitant non-target embolization and radiation of the non-tumorous liver parenchyma (NTLP) may lead to a localized and systemic inflammatory reaction [13,14]. Eventually, changes consistent with sinusoidal obstruction syndrome (SOS) and fibrosis can develop [15,16].
We hypothesized that the [ 18 F]FDG activity concentration in NTLP changes after radioembolization, leading to the misinterpretation and possible overestimation of the response using this activity concentration as a threshold in the PERCIST criteria. In this study, it was investigated whether the [ 18 F]FDG activity concentration in NTLP changes after radioembolization and whether alternative reference values can be used.  Table S1) and if a 90 Y-PET/CT was performed within 24 h of radioembolization for a regional dose calculation of the NTLP.

Study Population
The institutional medical ethics committee waived the need for informed consent for this retrospective review.

Radioembolization Treatment
After a pre-treatment hepatic arteriography to evaluate the vascular anatomy, a simulation was performed with the injection of 99m Tc-macroaggregated albumin (MAA) into the hepatic artery or arteries. Standard planar and SPECT images of 99m Tc-MAA were obtained for particle distribution assessments and to exclude relevant liver-lung shunting and extrahepatic depositions. In a separate session, 90 Y resin microspheres were injected into the hepatic artery or arteries after ensuring adequate catheter placement (similar to the 99m Tc-MAA injection position). The therapeutic activity of 90 Y resin microspheres was calculated according to the body surface area method.

[ 18 F]FDG-PET/CT Imaging Protocols
All [ 18 F]FDG-PET/CT scans at all time points were performed on the same PET/CTscanner (Biograph mCT, Siemens Healthcare, Erlangen, Germany). All patients fasted for at least 6 h prior to intravenous [ 18 F]FDG administration (2.0 MBq/kg) and each patient's blood glucose level was determined prior to the tracer injection (<11.1 mmol/L). The imaging parameters included a three-dimensional acquisition technique with a 216 mm field of view, 3 min per bed position, and ordered subset expectation maximization iterative reconstruction, including a Gaussian filter, 4 iterations, and 21 subsets. The measurements were performed on image reconstructions according to the EARL criteria [17].

90 Y-PET/CT Imaging Protocol
All 90 Y-PET/CT's were performed on the same PET/CT-scanner (Biograph mCT, Siemens Healthcare, Erlangen, Germany) within 24 h of treatment. The imaging parameters included a total acquisition time of 30 min for 2 bed positions, TrueX and time-of-flight reconstructions, and a reconstruction using 4 iterations with 21 subsets and a 5 mm full-with at half maximum Gaussian post-reconstruction filter. A low-dose CT scan was acquired for attenuation correction and as an anatomical reference.

Healthy Liver Parenchyma Analysis
All [ 18 F]FDG-PET/CT and 90 Y-PET/CT scans were analyzed using ROVER software (ABX, Radeberg, Germany). A 3 cm VOI in the right lobe of a normal liver is a robust indicator for the identification of the SUL mean [18] and is applied in the PERCIST criteria as a threshold parameter. To assess the NTLP activity concentration, the liver was separated into three regions, consistent with the three supplying arteries: the left hemiliver (i.e., the left hepatic artery (LHA)), the right hemiliver (i.e., the right hepatic artery (RHA)), and segment 4 (i.e., the middle hepatic artery (MHA) or segment 4 artery). Segment 4 was considered a separate region, as the MHA can be separately injected.
Spheric VOIs (3 cm in diameter) were placed in the NTLP in all three liver regions and in the spleen. If one of the regions was too diffusely involved in malignancy to accommodate a 3 cm VOI, no VOI was placed. In cases of (extended) right-sided hemihepatectomy, a second VOI was placed in the remaining liver tissue. Cilindrical VOIs with a 2 cm diameter were placed in the ascending, descending, and abdominal aortas, without inclusion of the aortic wall (to avoid elevated [ 18 F]FDG-uptake due to atherosclerosis).
The The SUL mean and SUL max were calculated for all VOIs. The same parameters were registered for the spleen, as well as the splenic volume. The splenic volume is known to increase after radioembolization; however, it is unknown whether the [ 18 F]FDG activity concentration changes are correlated to splenic volume changes [19].
In addition, all 90 Y-PET/CT scans were registered to the [ 18 F]FDG-PET/CT prior and the liver VOIs were co-registered onto the 90 Y-PET/CT, enabling a read-out of the absorbed dose (MBq/mL) in the same three regional VOIs. In a previous report, we validated the use of ROVER software for absorbed dose estimation on 90 Y-PET/CT [20].
All measurements on the [ 18 F]FDG-PET/CT prior were performed by one physician (MB; >10 years of experience).

Biochemical Changes
Laboratory examinations at the time of each PET/CT were noted, including liver function tests (total bilirubin, alkaline phosphatase, gamma-glutamyl transferase (GGT), aspartate aminotransferase (AST), alanine aminotransferase (ALT)) and coagulation parameters (thrombocytes, international normalized ratio (INR), partial thromboplastin time (PTT), activated partial thromboplastin time (APTT), and thrombin time (TT)). The Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 was used to grade biochemical toxicities. The presence of ≥2 CTCAE grade ≥1 for laboratory adverse events was regarded as clinically relevant biochemical toxicity.
Biochemical changes were correlated with changes in [ 18 F]FDG uptake in the NTLP.

Statistical Analysis
Descriptive statistics were used to explore baseline and treatment characteristics. The changes in SUL liver (the average of all VOIs per patient), SUL VOI,liver , SUL spleen , and SUL blood over time were assessed using linear mixed-effects regression models to account for clustered data. Nested models were compared using Akaike's Information Criterion. A model with a random slope only fitted the data best for all parameters. Time was used as a continuous independent variable. In the analyses of SUL liver and SUL VOI,liver , the injected activity (MBq) (patient-level) and the absorbed dose per VOI (VOI-level) were included as co-variables to adjust for possible confounding. The association between the parenchymal-absorbed dose and the change in SUL VOI,liver (represented as ∆SUL VOI,liver ) was also analyzed using a linear mixed-effects regression model with a random slope, with the absorbed dose as the continuous independent variable.
A logistic mixed-effects model was used to assess the association between the presence of clinically relevant biochemical toxicity (using the cut-off definition as described above) per time point and the change in SUL VOI,liver . The analysis was adjusted for response to therapy (coded as yes/no for progressive disease) as a possible confounder. All statistical analyses were done with R statistical software version 3.6.2 for Windows. We report here the effect estimates with associated 95% confidence intervals and corresponding two-sided p-values.  Table 1.   Table 2). The median absorbed dose per VOI was 31.3 Gy (range = 0.1-82.3 Gy). Overall, the median SULVOI,liver at baseline was 1.8 (range = 1.2-2.8). A significant increase in SULVOI,liver over time was found; the mean change in SULVOI,liver per month increase in time was 0.05 (95%CI 0.02-0.09) at p = 0.00088 ( Figure 2)( Table 3). The mean percent change in ΔSULVOI,liver for every Gy increase in dose was −0.04 (95%CI -0.22-0.14) at p = 0.67. At the patient level, the    Table 2). The median absorbed dose per VOI was 31.3 Gy (range = 0.1-82.3 Gy). Overall, the median SUL VOI,liver at baseline was 1.8 (range = 1.2-2.8). A significant increase in SUL VOI,liver over time was found; the mean change in SUL VOI,liver per month increase in time was 0.05 (95%CI 0.02-0.09) at p = 0.00088 ( Figure 2) (Table 3). The mean percent change in ∆SUL VOI,liver for every Gy increase in dose was −0.04 (95%CI -0.22-0.14) at p = 0.67. At the patient level, the median SUL VOI (average of all VOIs of the liver parenchyma) was 1.8 (range = 1.5-2.3). The mean change in SUL VOI per month increase in time was 0.05 (95%CI 0.01-0.08); p = 0.016 (Table 3).   Overall, the incidence of biochemical toxicity was low. CTCAE grade 3 toxicity was only found for bilirubin, GGT, and APTT (Table 4). PTT, INR, and TT were slightly prolonged in three, three, and one patient(s), respectively. The results of the logistic regression model showed a non-significant association between the clinically relevant  Overall, the incidence of biochemical toxicity was low. CTCAE grade 3 toxicity was only found for bilirubin, GGT, and APTT (Table 4). PTT, INR, and TT were slightly prolonged in three, three, and one patient(s), respectively. The results of the logistic regression model showed a non-significant association between the clinically relevant biochemical toxicity and change in SUL VOI (at the patient level) (95%CI 0.0006-333.29) at p = 0.90 (Table 5) (Figure 3). Table 4. Incidence rates of biochemical toxicity between baseline and three months after radioembolization according to CTCAE version 5.

Dependent variable
Odds ratio for clinically relevant biochemical toxicity for every month increase in SUL VOI,liver (95% CI); p-value  Table 4. Relation between SULVOI,liver and clinically relevant biochemical toxicity based on mixedeffects logistic regression analyses with SULVOI,liver as the independent variable.

Dependent variable
Odds ratio for clinically relevant biochemical toxicity for every month increase in SULVOI,liver (95% CI); p-value  For SUL blood , no significant increase or decrease over time was found; the mean change in SUL blood per month increase in time was −0.012 (95%CI −0.05-0.03) at p = 0.57 ( Figure 4A). SUL spleen was also stable over time; the mean change per month increase over time was 0.005 (95%CI −0.04-0.05) at p = 0.83 ( Figure 4B), while the splenic volume did not significantly increase over time. There was a 49 mL increase in volume per month (95%CI 32-65, p = 0.0011) ( Figure 4C).

Discussion
Our data showed that after 90 Y radioembolization, the [ 18 F]FDG activity concen (SULVOI liver) in the collateral targeted NTLP was significantly increased comp

Discussion
Our data showed that after 90 Y radioembolization, the [ 18 F]FDG activity concentration (SUL VOI liver ) in the collateral targeted NTLP was significantly increased compared to baseline. Although the SUL VOI liver increase was minimal, with a mean change in SUL VOI,liver per month increase over time of 0.05 with a median SUL VOI,liver at baseline of 1.8, this may influence quantitative assessments of liver activity concentrations as used in PERCIST. The influence on routine visual response assessments remains unclear, but will be minimal. Mild changes in [ 18 F]FDG activity concentration, however, could be noticed in our study visually ( Figure 5). baseline. Although the SULVOI liver increase was minimal, with a mean change in SULVOI,liver per month increase over time of 0.05 with a median SULVOI,liver at baseline of 1.8, this may influence quantitative assessments of liver activity concentrations as used in PERCIST. The influence on routine visual response assessments remains unclear, but will be minimal. Mild changes in [ 18 F]FDG activity concentration, however, could be noticed in our study visually ( Figure 5). Multiple reports on increased [ 18 F]FDG uptake in the liver after chemoradiation therapy for esophageal cancer have been published, with focal as well as diffuse patterns [16,[21][22][23][24]. However, only one publication has touched upon the [ 18 F]FDG uptake pattern The images are similarly scaled (0-7 SUV lean body mass ). One month after radioembolization, a reduced [ 18 F]FDG activity concentration can be seen in the metastases, but the background non-tumorous liver activity concentration has visually increased (SUL VOI,liver increase from 1.8 to 2.4 in the VOI right liver ).
Multiple reports on increased [ 18 F]FDG uptake in the liver after chemoradiation therapy for esophageal cancer have been published, with focal as well as diffuse patterns [16,[21][22][23][24]. However, only one publication has touched upon the [ 18 F]FDG uptake pattern in normal liver parenchyma 1 month after radioembolization, with no changes compared to baseline [25]. Unfortunately, additional data were lacking and no correlation was made with the absorbed dose.
Remarkably, we did not find an association between the median absorbed dose and the change in SUL VOI,liver . In contrast, Nakahara et al. reported a direct visual correspondence to the radiation dose distribution and the pattern of increased [ 18 F]FDG uptake in the liver in a patient after chemoradiation therapy for esophageal cancer [24]. The lack of association in our data may have been due to the limited patient numbers and the limited spread in absorbed doses between the VOIs.
The explanation for the increased [ 18 F]FDG liver activity concentration after radioembolization is largely unknown. Local inflammation seems plausible, although this is not confirmed in the existing (although scarce) histopathology literature [15,[26][27][28]. An in vitro assessment of ovarian cancer cell lines after radiotherapy showed swelling and increased [ 3 H]FDG uptake in the surviving cells [29]; a similar phenomenon may be present in hepatocytes after radioembolization. However, the most plausible explanation may be increased [ 18 F]FDG uptake due to sinusoidal obstruction syndrome (SOS). Several authors reported signs of SOS developing within the first 3 months after radioembolization [15,27,28]. As shown by Kim et al. in 35 patients with SOS after platin-based chemotherapy, SOS can lead to a significantly increased [ 18 F]FDG liver activity concentration [30]. The possible explanation they offer is passive [ 18 F]FDG tracer stasis due to endothelial cell injury and peliotic changes. The increase in SUL liver in their study (12%) was considerably higher than in our data, yet with similar to our findings; the activity concentration in the blood pool remained unchanged, unlike the splenic activity concentration (which increased significantly in moderate to severe cases of SOS in their study). As the blood pool uptake seems unchanged at different time points after injection, in different patient populations and even after chemotherapy, it may be the best reference value for patients treated with radioembolization [4,[6][7][8][9]30,31].
Although PERCIST aims to unify response assessments with [ 18 F]FDG-PET, in our study 30.8% of the patients with [ 18 F]FDG-PET/CT scans at baseline and follow-up were excluded based on the PERCIST timing and scan criteria (apart from the liver SUL criterion). This is consistent with most publications, which show mean exclusion percentages of 22.6-38.3% for PERCIST response assessments, due to the strict criteria for scan comparability (Supplementary Table S1) [9,10,32]. The major restrictions include the difference in post-injection scanning time (<15 min) and the minimum of 50 min needed for postinjection scanning. Unfortunately, these are significant factors in SUV measurements and major logistical problems for most nuclear medicine departments. As shown in several test-retest studies, the interval between injection and acquisition is a significant parameter for underestimating liver activity [31][32][33][34]. This may be explained by the abundance of the enzyme glucose-6-phoshphatase in the liver, causing continuous glycolysis and a decrease in FDG retention [35]. Thus, it is essential to minimalize the differences in FDG uptake time between scans to ensure a consistent threshold for PET-based response assessments (as used in PERCIST, the Deauville criteria, etc.). The post-injection times between [ 18 F]FDG-PET/CT scans in our study were not significantly different (Table 2). However, in 5/26 of the right liver VOIs, the SUL VOI, liver showed a ≥0.3 SUL unit change, violating the scan restrictions according to the PERCIST criteria. In these cases, the radioembolization-induced increased [ 18 F]FDG liver activity concentration would have resulted in exclusion. However, these minimal changes in SUL VOI will probably not be clinically relevant in routine visual response assessments. With the increased use of the modified PERCIST criteria, the reference threshold of the VOI liver is lowered, and the use of a more robust reference value in clinical studies is advisable [36].
A comparison of [ 18 F]FDG liver activity concentrations with earlier reported results is difficult, due to the differences in the quantification or correction of the [ 18 F]FDG uptake, i.e., the use of ROIs or VOIs; the use of SUV max , SUV mean , or the total lesion/liver glycolysis (TLG); and correction for the body surface area (BSA) vs. body weight vs. lean body mass (LBM). Only the LBM-corrected SUV (SUL) is not significantly affected by body weight, contrary to the BSA-corrected or body-weight-corrected SUV. The calculations are not affected by age, blood glucose, diabetes, or gender [3,4,31]. Therefore, in our study, SUL was chosen for all measurements, consistent with the PERCIST criteria.
There were a few limitations to this study, the foremost of which was the small patient number with consistently acquired PET/CT scans according to the PERCIST guidelines. A patient cohort with only lobar treatments, enabling an in-patient comparison with nontreated NTLP, would be better by avoiding inter-patient biological differences. Additionally, a comparison of patients treated with different commercially available microspheres (with their differing specific activity levels) could help to further evaluate the role of [ 18 F]FDG PET/CT as a follow-up tool after radioembolization. Given the expanding use of both external beam radiotherapy and radioembolization and the increasing use of PET/CT for response assessments, further research is required.

Conclusions
The [ 18 F]FDG liver activity concentration in the background liver (SUL VOI ) is mildly but significantly increased after radioembolization; therefore, this can interfere with the use of the background liver results firstly as a threshold for therapy response and secondly as a comparability criterion in the PERCIST criteria. An alternative and more consistent threshold is the [ 18 F]FDG activity concentration in the blood pool (SUL blood ).

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/diagnostics12102518/s1, Table S1: PERCIST criteria. Institutional Review Board Statement: Ethical review and approval were waived for this study due to the retrospective nature of this study and the informed consent obtained from all patients to use their imaging and clinical data (as participant in other studies on radioembolization treatments).

Informed Consent Statement:
Written informed consent was waived by the Institutional Review Board.

Data Availability Statement:
The data presented in this study are available on reasonable request from the corresponding author.

Conflicts of Interest:
The authors of this manuscript declare relationships with the following companies: BTG/Boston Scientific, Terumo, and Quirem. A. Braat is a speaker for BTG/Boston Scientific and Terumo. M. Lam receives research support from Terumo, Quirem, and Boston Scientific, and is a consultant on their behalf. The department of Radiology and Nuclear Medicine of the University Medical Center Utrecht receives royalty payments from Quirem. Their radioembolization products were, however, not used in this article. The other authors of this manuscript (C. van Roekel, M.N.G.J.A. Braat) declare no relationships with any companies whose products or services may be related to the subject matter of the article.